search

Denys88 / rl_games

RL implementations
MIT License
895 stars 150 forks source link

Fix multi-gpu training with horovod. #119

Closed ArthurAllshire closed 2 years ago

Denys88 commented 2 years ago

@ArthurAllshire if it is scaler for mixed precision I believe we can write wrapper and make sure it stores everything as pytorch tensors.